Hi, I just install the program BOINC Manager and I searched and I could not find a way to make GPU to calculate. From what I read in forum I understood that there are tasks for GPU's and CPU's. I just want to know is there any way to make my VGA to start working on the CPU problems. I have GTS450 and I have used another program made by NVIDIA and there was just one mark and I could choose CPU or GPU. In BOINC Manager I set to use the GPU always, to use it when I am using the PC, but the GPU is still staying idle. I whant to know is there another kind of options or I am doing something wrong? And how to know is the task for the CPU or the GPU?

---

I'm not running the generic SETI programs anymore, but the applications from Lunatics http://lunatics.kwsn.net/ . I've noticed that you can see in your results whether it was a CPU task or a GPU task (or maybe both). I think the main stream SETI apps have some NVIDIA support, but others could speak better to that than me since I have an ATI card. I've noticed from the event log that BOINC will request CPU and/or GPU tasks. The real question in my mind is why is there a difference between a CPU and a GPU task? I would think that SETI just provides the data (and their generic apps with the exception of NVIDIA support) and processes the results that come back. It seems to me that it ought to be a (data in)-(black box compute)-data out sort of model. In other words let the app figure out how to process the data and use whatever resources it has access to. (CPU/GPU/FPGA...) Is there some sort of preprocessing of data done before WUs are sent out so that it can be determined whether or not it's a CPU or GPU or CPU/GPU WU? I'm sure somebody out there can explain this to me. Interestingly enough I found a roughly 2.5 times speed increase using the Lunatics apps over generic SETI and was very happy, until I downloaded GPU-Z and found out my GPU wasn't even being used. I put a load on it with some OpenCL apps and verified the GPU was working correctly. Either way the Lunatics apps definitely perform better and you'll have much more support with NVIDIA than with ATI. I hope that didn't sound angry, it wasn't. There's just a lot of work to do and much that has to be figured out.

Thanks for the answer. I noticed that in application tab in different project is written that is using CUDA (for example PrimeGrid). If it is not written then you use only CPU.

---

The task files (e.g. 20ap11ad.22038.9883.8.10.220) are the same not matter which app/device will process them.

Meaning if you look inside such task file you will not see anything that can inform you what device on your computer will process the task.

A task file (for SETI@home Enhanced) look like this inside:

<workunit>
<workunit_header>
  <name>20ap11ad.22038.9883.8.10.220</name>
  <group_info>
    <tape_info>
      <name>20ap11ad</name>
      <start_time>2455672.1802263</start_time>
      <last_block_time>2455672.1802263</last_block_time>
      <last_block_done>9883</last_block_done>
      <missed>0</missed>
      <tape_quality>0</tape_quality>
      <beam>5</beam>
    </tape_info>
    <name>20ap11ad.22038.9883.8.10</name>
    <data_desc>
      <start_ra>2.0868110548832</start_ra>
      <start_dec>32.724148238801</start_dec>
      <end_ra>2.11669460206</end_ra>
      <end_dec>32.724112185166</end_dec>
      <true_angle_range>0.37710791792417</true_angle_range>
      <time_recorded>Wed Apr 20 16:19:31 2011</time_recorded>
...
...
...
B+_*#;$?GQ:+$O,*'8U+8Y%7 [4U["*%CMWB<FL%DFRRTS>%65J\.9I-CIV^Q,21
'4MZ"QSZ0^E);"\6T8.DS<#,NIT>2>$)LBWD=VYG9EUPJ5$B,8]1)0NIU*=^?3R>
KF;$[;BPX0;)9@QA?<)K*   
</data>
</workunit>   

BOINC decides which device needs work and how much and asks the servers to send tasks.
The tasks itself are the same but some info for them (e.g. filename, md5_cksum, for which device/app they are sent) goes to you in sched_reply_setiathome.berkeley.edu.xml

This task-describing info is then put by BOINC in client_state.xml

So the task files are the same, only the additional info describing them is different.
This additional info (e.g. for CPU or GPU) is generated at the time of your communication with the servers and not before.

The same task can be "labeled" in many different ways depending what "you" ask for.
( ;) Like some fake-beverage dealer can put the same whisky in different brand/label bottles)


In BOINC Manager -> Tasks tab -> Application column you can tell the difference (the target device) by version:
- setiathome_enhanced 6.03 (or <6.03) is for CPU
- setiathome_enhanced 6.08 (or >6.08) is for GPU


Look here for the (standard) versions:
http://setiathome.berkeley.edu/apps.php

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Am I mistaking or is it not as easy to just go in to the "Tools -> Computing preference -> Use GPU" ?

Should that not make me use my GPU? Just installed the BOINC Manager an hour ago for the first time, thats why I am asking :)

Best regards,
Davoud

You have ATI/AMD Radeon - SETI still do not have standard Application for ATI/AMD GPUs:
http://setiathome.berkeley.edu/apps.php

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

You have ATI/AMD Radeon - SETI still do not have standard Application for ATI/AMD GPUs:
http://setiathome.berkeley.edu/apps.php

In other words, I cant use my GPU?

You have ATI/AMD Radeon - SETI still do not have standard Application for ATI/AMD GPUs:
http://setiathome.berkeley.edu/apps.php

In other words, I cant use my GPU?

Not unless you use a third-party science application that supports your GPU [edit] or you connect to a project that does support ATi GPUs natively.

If you choose to run the 3rd party app for seti, You take responsibility for keeping it updated. You will also find your GPU will run hotter using the 3rd party app. This means you'll need to manually adjust your GPU fan speed. The 3rd party app has a portion that will make your CPU WU's finish in about 1/2 their normal run times.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope